Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Accurate and Low-Parameter Machine Learning Architecture for Next Location Prediction

Calvin Jary, Nafiseh Kahani

TL;DR

This work tackles next location prediction in mobile networks with a demand for energy-efficient, deployment-friendly models. It proposes a lean, two-layer RNN architecture with embeddings and a softmax output, optimized through extensive hyperparameter exploration on a city-scale Changchun mobility dataset. The approach achieves comparable or better accuracy while drastically reducing model size (from ~791 MB to ~8 MB) and parameter count (from ~202M to ~2M), and it accelerates training (~4x) with far lower GPU memory usage. The results demonstrate strong edge-deployability and generalization potential across city-scale mobility scenarios, paving the way for practical, scalable location-prediction systems on base stations and edge devices.

Abstract

Next location prediction is a discipline that involves predicting a users next location. Its applications include resource allocation, quality of service, energy efficiency, and traffic management. This paper proposes an energy-efficient, small, and low parameter machine learning (ML) architecture for accurate next location prediction, deployable on modest base stations and edge devices. To accomplish this we ran a hundred hyperparameter experiments on the full human mobility patterns of an entire city, to determine an exact ML architecture that reached a plateau of accuracy with the least amount of model parameters. We successfully achieved a reduction in the number of model parameters within published ML architectures from 202 million down to 2 million. This reduced the total size of the model parameters from 791 MB down to 8 MB. Additionally, this decreased the training time by a factor of four, the amount of graphics processing unit (GPU) memory needed for training by a factor of twenty, and the overall accuracy was increased from 80.16% to 82.54%. This improvement allows for modest base stations and edge devices which do not have a large amount of memory or storage, to deploy and utilize the proposed ML architecture for next location prediction.

An Accurate and Low-Parameter Machine Learning Architecture for Next Location Prediction

TL;DR

This work tackles next location prediction in mobile networks with a demand for energy-efficient, deployment-friendly models. It proposes a lean, two-layer RNN architecture with embeddings and a softmax output, optimized through extensive hyperparameter exploration on a city-scale Changchun mobility dataset. The approach achieves comparable or better accuracy while drastically reducing model size (from ~791 MB to ~8 MB) and parameter count (from ~202M to ~2M), and it accelerates training (~4x) with far lower GPU memory usage. The results demonstrate strong edge-deployability and generalization potential across city-scale mobility scenarios, paving the way for practical, scalable location-prediction systems on base stations and edge devices.

Abstract

Next location prediction is a discipline that involves predicting a users next location. Its applications include resource allocation, quality of service, energy efficiency, and traffic management. This paper proposes an energy-efficient, small, and low parameter machine learning (ML) architecture for accurate next location prediction, deployable on modest base stations and edge devices. To accomplish this we ran a hundred hyperparameter experiments on the full human mobility patterns of an entire city, to determine an exact ML architecture that reached a plateau of accuracy with the least amount of model parameters. We successfully achieved a reduction in the number of model parameters within published ML architectures from 202 million down to 2 million. This reduced the total size of the model parameters from 791 MB down to 8 MB. Additionally, this decreased the training time by a factor of four, the amount of graphics processing unit (GPU) memory needed for training by a factor of twenty, and the overall accuracy was increased from 80.16% to 82.54%. This improvement allows for modest base stations and edge devices which do not have a large amount of memory or storage, to deploy and utilize the proposed ML architecture for next location prediction.
Paper Structure (17 sections, 6 figures, 2 tables)

This paper contains 17 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background and Related Works
  3. Approach
  4. Input Size
  5. Embedding Dimensions
  6. RNN Block
  7. Hidden Nodes
  8. Number of Layers
  9. Linear Layer and Softmax Function
  10. Validation
  11. Research Questions
  12. Dataset
  13. Experiment Design
  14. Results and Discussion
  15. RQ1. A simpler ML architecture in the context of next location prediction.
  16. ...and 2 more sections

Figures (6)

  • Figure 1: High-level architecture of the proposed ML model
  • Figure 2: Structure of the dataset
  • Figure 3: Model accuracy and its relation to window size (total input locations).
  • Figure 4: Model accuracy and its relation to the amount of hidden nodes.
  • Figure 5: Model accuracy and its relation to the number of embedding dimensions (total learned features per location).
  • ...and 1 more figures